From Personalized Learning To Social-Emotional Development: Applications for AI in Education

The integration of artificial intelligence (AI) and large language models (LLMs) into educational research is revolutionizing the way we approach teaching and learning. Researchers are seeking ways to bring AI into practice to benefit education and society. For instance, research projects which bridge academia, industry, government, and civil society are central to creating applicable knowledge and to delivering contextually-relevant innovations. 

The MIT Jameel World Education Lab is supporting MIT researchers that are driving such change through collaborative approaches to research, teaching, and engagement. Four innovative projects funded by J-WEL grants in Education Innovation are at the forefront of this transformation, each leveraging AI and LLMs to address unique educational challenges and improve learner outcomes. 

Enhancing Active Learning in the Classroom 

Rea Lavi at the Department of Aeronautics and Astronautics is developing a web-based platform, SIDAI, and a chatbot teaching assistant named Sid that will assist in active learning and provide personalized feedback to students, aiming to enhance teaching and learning experiences. Active learning, which involves engaging students in activities that promote analysis, synthesis, and evaluation of class content, is known to enhance critical thinking and problem-solving skills. However, its implementation can be challenging due to large class sizes and the need for extensive preparation. This challenge motivated Lavi to explore the use of generative AI to remedy some of the hindrances with traditional pedagogical approaches.

SIDAI addresses these challenges by using generative AI to create a scalable platform that provides personalized feedback to students. The platform guides both instructors and students through active learning processes, making it easier to implement these techniques in large classes. The integration of AI allows for the automation of routine tasks and the provision of real-time, individualized feedback, which would be difficult to achieve manually.

Lavi says that success will look like “high levels of satisfaction, usability, and perceived efficacy for teaching and learning on a survey of instructors and survey of students.”

He continues, “we are now working on developing the second version of the platform (which includes the Sid chatbot) and plan to pilot the platform in spring. We have secured pilots in undergraduate STEM courses with Aalborg University and with Purdue University.”

Personalized Learning Through Physiological Sensing

Nataliya Kosmyna and Pattie Maes at the MIT Media Lab are developing an adaptive learning platform, NeuroChat, that uses brain-sensing biofeedback to personalize learning experiences based on a student's cognitive state. By measuring engagement through wearable EEG headbands, NeuroChat can adjust the complexity of explanations and the use of figurative language to maintain student interest and improve learning outcomes.

The preliminary findings from NeuroChat's first study indicate that this approach positively impacts engagement and learning outcomes. By providing additional context about the user's cognitive state, NeuroChat enhances the interaction between students and AI tutors, making learning more effective and personalized. The innovation, which pairs wearable tech with an online platform, brings a sense of novelty to the human-AI interaction

Kosmyna says that the next step in their study will be outside of the lab, meaning that “anyone who uses the internet will be able to try and use NeuroChat. We want to know what people all around the world think about it, how and why they use it, and learn more about their use cases to adapt this system to the needs of humans.”

Bridging the Vocabulary Divide through Personalized Tutoring

Ola Ozernov-Palchik, Fabio Catania, Satra Ghosh, and John D.E. Gabrieli at the McGovern Institute for Brain Research are leveraging AI and LLMs to address the vocabulary knowledge gap in young children. Vocabulary knowledge is a critical predictor of reading achievement and overall academic success, yet significant gaps persist due to socioeconomic disparities. 

The team is developing a speech-based LLM-empowered conversational tutor to enhance vocabulary knowledge in third and fourth graders from diverse socioeconomic backgrounds. By leveraging AI, the tutor can provide personalized and scalable vocabulary instruction, adapting to each child's needs and progress. 

Ozernov-Palchik said that AI “enables capabilities that would otherwise be impossible. AI-based modeling is essential for automatic child speech processing, allowing our avatar to understand and engage with children in natural, interactive conversations. Without these tools, the avatar would lack the ability to process and respond to a child’s speech effectively.”

The project's ultimate goal is to “contribute to a meaningful improvement in national literacy achievement scores, with a particular emphasis on reducing the literacy gap between children from disadvantaged backgrounds and their peers,” continued Ozernov-Palchik.

Supporting Social-Emotional Learning in Refugee Children

Sharifa Alghowinem and Hae Won Park at the MIT Media Lab are developing a social robot platform tailored for Arabic-speaking refugee children. Equipped with culturally sensitive design and state-of-the-art algorithms, the platform aims to enhance reading, vocabulary, and social-emotional learning (SEL) in this population of children through culturally sensitive interactions.

The platform features interactive Arabic storybooks and emotional conversation simulation scenarios that the research team developed using customized Arabic automatic speech recognition, text-to-speech, and interactive applications to provide engaging and relevant educational content. 

The team is partnering with NGOs including We Love Reading, ANERA, iLearn, JMAP, and Education Above All to pave the way for experimental setups and deployment preparations, ensuring that the content is culturally appropriate and addresses the specific needs of refugee children. 

Alghowinem says that “AI is central to our project's success, enabling us to deliver personalized and culturally sensitive learning experiences.” 

The Future of AI in Education

These early project outcomes demonstrate the transformative potential of AI for education, and have the potential to contribute to a more inclusive and effective educational system. As AI technology continues to evolve, we expect to see even more innovative applications that individualize learner and educator feedback, adapt to learners' needs, and scale educational interventions. 

J-WEL is proud to support MIT research that extends its impact beyond our campus, including benefitting collaborators outside of academia, outside of the U.S., and in more marginalized learning communities. We look forward to tracking and sharing the outcomes of this work over the coming year.

‍